Outgoing devserver traffic is not very balanced between devservers |
||||
Issue descriptionUsing outgoing network traffic from devservers as a proxy for how much data is being served shows that load is not very balanced between devservers: http://shortn/_qwdFrCEu8B The difference appears to be 1.5-2x at any given time and is present even if the window is increased from 2h.
,
Jul 18 2017
The imbalance does not bother me, by itself, unless we are also overloading some of the devservers (which we often do). There is a general belief that we need to improve the mechanism used to resolve which devserver to use for a given test, but I don't know if there are existing bugs for that.
,
Sep 26 2017
Richard's changes might have addressed this. It's worth looking at metrics to see if that's true.
,
Jun 8 2018
Hi, this bug has not been updated recently. Please acknowledge the bug and provide status within two weeks (6/22/2018), or the bug will be archived. Thank you.
,
Aug 2
,
Aug 3
I've done nothing that would make this better. From what I've seen, a small but measurable fraction of provisioning tasks do fail while interacting with the devservers in ways that are conveniently (though possibly not always accurately) described as "the devserver was too busy." |
||||
►
Sign in to add a comment |
||||
Comment 1 by dgarr...@chromium.org
, Apr 13 2017